Re: [PATCH v3 1/5] dt-bindings: pci: Add ARTPEC-8 PCIe controller

[Krzysztof Kozlowski]

On 21/06/2022 09:42, Wangseok Lee wrote:
>>>  
>>>  samsung,syscon-bus-s-fsys:
>>>    description:
>>>      Phandle to bus-s path of fsys block, this register
>>>      are used for enabling bus-s.
>>>    $ref: /schemas/types.yaml#/definitions/phandle
>>>  
>>>  samsung,syscon-bus-p-fsys:
>>>    description:
>>>      Phandle to bus-p path of fsys block, this register
>>>      are used for enabling bus-p.
>>>    $ref: /schemas/types.yaml#/definitions/phandle
>>
>> This two look unspecific and hacky workaround for missing drivers. Looks
>> like instead of implementing interconnect or clock driver, you decided
>> to poke some other registers. Why this cannot be an interconnect driver?
>>
>>
> 
> bus-s, bus-p is a register that exists in the sysreg of the fsys block.
> It is the same block as "fsys-sysreg" but is separated separately in
> hardware.

Two points here:
1. If it is in FSYS, why it cannot be accessed with samsung,fsys-sysreg?
2. If it is only register, shuld be described like this. You must
describe item:
https://elixir.bootlin.com/linux/v5.18-rc1/source/Documentation/devicetree/bindings/soc/samsung/exynos-usi.yaml#L42

> So, get resource is performed separately from "fsys-sysreg".
> They set pcie slave, dbi related control settings,
> naming "bus-x" seems to be interconnect.
> I will add this description to property.
> I don't think it need to use the interconnect driver,
> so please let me know your opinion.

Please document both in the bindings and in the driver usage of this
register. Writing there "0" or "1" is not enough. If the documentation
is good, I am fine with it. If the explanation is obfuscated/not
sufficient, it will look like avoiding to implement a driver, which I
don't want to accept.

Best regards,
Krzysztof